DESIGN DEVELOPMENT TOOLS CHAPTER 13: DESIGN DEVELOPMENT TOOLS INTRODUCTION 13.1 SECTION 13.1: SIMULATION 13.3 SPICE 13.3 MACROMODEL VS. MICROMODEL 13.4 THE ADSPICE OP AMP MICROMODELS 13.5 INPUT AND GAIN/POLE STAGES 13.6 FREQUENCY SHAPING STAGES 13.7 MACROMODEL OUTPUT STAGES 13.8 MODEL TRANSIENT RESPONSE 13.9 THE NOISE MODEL 13.10 CURRENT FEEDBACK MODLES 13.11 SIMULATION MUST NOT REPLACE BREADBOARDING 13.13 SIMULATION IS A TOOL TO BE USED WISELY 13.14 KNOW THE MODELS 13.14 UNDERSTANDING PCB PARASITICS 13.14 SIMULATION SPEEDS THE DESIGN CYCLE 13.16 SPICE SUPPORT 13.17 MODEL SUPPORT 13.17 IBIS MODELS 13.17 SABER MODELS 13.17 ADIsimADC 13.18 BEHAVIORAL VS. BIT EXACT 13.18 MODEL VS. HARDWARE 13.18 WHAT IS IMPORTANT TO MODEL? 13.19 GAIN, OFFSET, AND DC LIEARITY 13.19 SAMPLE RATE AND BANDWIDTH 13.21 DISTORTION, BOTH STATIC AND DYNAMIC 13.22 JITTER 13.24 LATENCY 13.25 ADIsimPLL 13.26 REFERENCES 13.31 SECTION 13.2: ON-LINE TOOLS AND WIZARD 13.33 SIMPLE CALCULATORS 13.33 CONFIGURTION ASSISTANTS 13.46 BASIC LINEAR DESIGN DESIGN WIZARDS 13.58 PHOTODIODE WIZARD 13.58 ANALOG FILTER WIZARD 13.61 SUMMARY 13.68 SECTION 13.3: EVALUATION BOARDS AND PROTOTYPING 13.69 EVALUATION BOARDS 13.69 GENERAL-PURPOSE EVALUATION BOARD 13.69 DEDICATED OP AMP EVALUATION BOARDS 13.70 DATA CONVERTER EVALUATION BOARDS 13.72 HIGH SPEED FIFO EVALUATION BOARD SYSTEM 13.74 FIFO BOARD THERORY OF OPERATION 13.75 CLOVKING DESRIPTION 13.76 CLOCKING WITH INTERLEAVED DATA 13.78 CONTROLLER FOR PRECISION ADCs 13.79 HARDWARE DESCRIOTION 13.80 COMMUNICATIONS 13.80 POWER SUPPLIES 13.80 OUTPUT CONNECTOR 13.80 SOFTWARE 13.81 PROTOTYPING 13.82 DEADBUG PROTOTYPING 13.82 SOLDER MOUNT PROTOTYPING 13.84 MILLED PCB PROTOTYPING 13.86 BEWARE OF SOCKETS 13.87 SOME ADDITIONAL PROTOTYPING HINTS 13.88 FULL PROTOTYPE BOARD 13.89 SUMMARY 13.90 REFERENCES 13.91 DESIGN DEVELOPMENT TOOLS INTRODUCTION CHAPTER 13: DESIGN DEVELOPMENT TOOLS Introduction There are several tools available to help in the design and verification of a design. From software that helps in the part selection to circuit simulation and error budget analysis, these tools make the process of design faster, easier, and more exact. The first section covers the simulation software. Primary among these is Spice. Originally developed for analysis of linear circuits for the integrated circuit industry (Spice is an acronym for Simulation Program with Integrated Circuit Emphasis), it has expanded to become the primary linear simulation engine. It has become available in many versions, such as Pspice®, Hspice®, as well as others. The original Spice was developed in the mid- 1970s. It originally ran on mainframes, obviously without the benefit of a graphical user interface (GUI). Most of the proprietary versions use front end and back end processing to make the human interface easier. Some add other features, such as Monte-Carlo analysis. All handle the basic Spice functions the same, however. At Analog Devices, we have settled on Pspice as our standard, but we use only the basic SPICE2-G functions. This makes the files more transportable. For very large circuits, such as a data converter, Spice becomes increasingly harder to use. For this reason, we shift from component level modeling to behavioral modeling. An example of this is the ADIsimADC program, which is also discussed in this section. In the second section we discuss the family of on-line tools developed by Analog Devices. These range from relatively simple tools, such as the settling time calculator to “wizards,” such as the filter design wizard. Wizards help design the circuits as well as assist in part selection. No matter how much simulation and design tools may help with the design of the circuit, there is absolutely no substitute for actually building the circuit. The last section discusses evaluation boards and prototyping techniques. 13.1 BASIC LINEAR DESIGN 13.2 DESIGN DEVELOPMENT TOOLS SIMULATION SECTION 13.1: SIMULATION Spice In the past decade, circuit simulation has taken on an increasingly important role within analog circuit design. The most popular simulation tool for this is Spice, which is available in multiple forms for various computer platforms (see References 1 and 2). However, to achieve meaningful simulation results, designers need accurate models of many system components. The most critical of these are realistic models for ICs, the active devices that drive modern designs. In the early 1990s, Analog Devices developed an advanced op amp Spice model, which is in fact still in use today (see References 3 and 4). Within this innovative open amplifier architecture, gain and phase response can be fully modeled, enabling designers to accurately predict ac, dc, and transient performance behavior. This modeling methodology has also been extended to include other devices such as in-amps, voltage references, and analog multipliers. The popularity of Spice simulation has led to many op amp macromodel releases, which (ideally) software-mimic amplifier electrical performance. With numerous models available, several confusions are possible. There may be uncertainty as to what is/isn’t modeled, plus a basic question of model accuracy. All of these points are important, in order to place confidence in simulation results. So, verification of a model is important, checking it by comparison to the actual device performance conditions, before trusting it for serious designs. Of course, a successful first design step using an accurate op amp model by itself doesn’t necessarily guarantee totally valid simulations. A simulation based on incomplete information has limited value. All parts of a target circuit should be modeled, including the surrounding passive components, various parasitic effects, and temperature changes. Then, the circuit needs to be verified in the lab, by breadboarding and prototyping. A breadboard circuit is a quickly executed mockup of a circuit design using a semi- permanent lab platform, i.e., one which is less than final physical form. It is intended to show real performance, but without the total physical environment. A good breadboard can often reveal behavior not predicted by Spice, either because of an incomplete model, external circuit parasitics, or numerous other reasons. However, by using Spice along with intelligent breadboarding techniques, a circuit can be efficiently and quickly designed with reasonably good assurance of working properly on a prototype version, or even a final PCB. The following prototype phase is just one step removed from a final PCB, and may in fact be an actual test PCB, with nearly all design components incorporated, and with close to full performance. The breadboard/prototype design steps are closely allied to simulation, usually following it in the overall design process. These are more fully discussed in subsequent sections. 13.3 BASIC LINEAR DESIGN Macromodel vs. Micromodel The distinction between macromodel and micromodel is often unclear. A micromodel uses the actual transistor level and other Spice models of an IC device, with all active and passive parts fully characterized according to the manufacturing process. In differentiating this type of model from a macromodel, some authors use the term device level model to describe the resulting overall op amp model. Typically, a micromodel is used in the actual design process of an IC. METHODOLOGY ADVANTAGES DISADVANTAGES IDEAL ELEMENTS FAST SIMULATION MAY NOT MODEL ALL MACROMODEL MODEL DEVICE TIME, CHARACTERISTICS BEHAVIOR EASILY MODIFIED FULLY MOST COMPLETE SLOW SIMULATION, MICROMODEL CHARACTERIZED MODEL CONVERGENCE TRANSISTOR LEVEL DIFFICULTY, CIRCUIT NON-AVAILABILITY Figure 13.1: Differentiating the Macromodel and Micromodel A macromodel goes another route in emulating op amp performance. Taking into consideration final device performance, it uses ideal native Spice elements to model observed behavior—as many as necessary. In developing a macromodel, a real device is measured in terms of lab and data sheet performance, and the macromodel is adjusted to match this behavior. Some aspects of performance may be sacrificed in doing this. Figure 13.1 compares the major pros and cons between macromodels and micromodels. There are advantages and disadvantages to both approaches. A micromodel can give a complete and accurate model of op amp circuit behavior under almost all conditions. But, because of a large number of transistors and diodes with nonlinear junctions, simulation time is very long. Of course, manufacturers are also reluctant to release such models, since they contain proprietary information. And, even though all transistors may be included, this isn’t a guarantee of total accuracy, as the transistor models themselves don’t cover all operational regions precisely. Furthermore, with a high node count, Spice can have convergence difficulties, causing a failed simulation. This point would make a micromodel virtually useless for multiple amplifier active filters, for example. On the other hand, a carefully developed macromodel can produce both accurate results and simulation time savings. In more advanced macromodels such as the ADSpice model described below, transient and ac device performance can be closely replicated. Op amp nonlinear behavior can also be included, such as output voltage and current swing limits. However, because these macromodels are still simplifications of real devices, all nonlinearities aren’t modeled. For example, not all ADSpice models include common- mode input voltage range, or noise (while more recent ones do). Typically, in model 13.4 DESIGN DEVELOPMENT TOOLS SIMULATION development parameters are optimized as may be critical to the intended application— for example, ac and transient response. Including every possible characteristic could lead to cumbersome macromodels that may even have convergence problems. Thus, ADSpice macromodels include those op amp behavior characteristics critical to intended performance for normal operating conditions, but not necessarily all nonlinear behavior. The ADSpice Op Amp Macromodels The basic ADSpice model was developed as an op amp macromodeling advance, and as an improved design tool for more accurate application circuit simulations.